The purpose of our study is to address the challenge of evaluating carcinogenic effects at low levels of exposure to carcinogens. We examine the shape of the dose-response relationship between tobacco smoking and cancers of the bladder and lung, and the implications for the evaluation of the effects of exposure at lower levels, for example, environmental tobacco smoke (ETS).
DOSE-RESPONSE RELATIONSHIPS IN TOBACCO CARCINOGENESIS: BLADDER AND LUNG CANCERThe fact that low doses of tobacco may have a carcinogenic effect proportionally greater than high doses is suggested by the study of dose-response relationships. A previous study, based on the re-analysis of a multicenter case-control study on lung cancer and of several studies on bladder cancer, 1 suggested that the dose-response relationship between cigarette smoking and cancer risk tends to level off after a dose of approximately 20 -25 cigarettes/day. A few explanations were put forward to explain leveling off, including bias (less accurate reporting by heavy smokers with cancer), lower inhalation at high doses or genetic heterogeneity of the population, with a "depletion of susceptibles" at low-dose levels. To test the latter hypothesis, we have reviewed the recent studies on smoking and bladder cancer, a type of tumour that is particularly well studied from a biochemical-molecular point of view (Tables I and II).We have identified all the cohort or case-control studies on smoking and bladder cancer published from 1985 to 2002 (those published before were reviewed in the IARC Monograph on tobacco smoking 2 ). We have considered men only because data for women tended to be unstable. We have extracted dose-response data, showing odds ratios and 95% confidence intervals whenever possible. Table I shows the results of case-control studies, and Table II those of cohort studies. Virtually all studies, except 1 case-control (Momas et al., 1994) and 2 cohort studies (Engeland et al., 1996;Tulinius et al., 1997), show a leveling off after 20 -30 cigarettes/day. Engeland et al. (1996) and Tulinius et al. (1997) show an attenuation of the slope but not a clear leveling off. The consistency between the 2 different designs suggests that we are observing a real phenomenon and not an artifact because different types of bias occur in case-control and in cohort investigations. In fact, only the former design is prone to retrospective recall bias, with underestimation of heavy consumption by cancer cases.There are other examples in the literature in which the shape of the dose-response relationship levels off. The relationship between TCDD (dioxin) exposure (measured in the plasma of the exposed workers) and total mortality from cancer 3 shows a plateauing of the curve, i.e., an effect that is proportionally greater at lower levels of exposure. Other examples are liver tumours and vinyl chloride in rats 4 or lung cancer and arsenic in humans. 5 However, it should be noted that these are heterogeneous situations, concerning both mutagens like vinyl chloride and chemicals like...